[PATCH 2/3] drm/fb_cma_helper: Add panic handling

[Laurent Pinchart]

Hi Noralf,

Thank you for the patch.

On Sunday 11 Sep 2016 20:47:41 Noralf Trønnes wrote:
> This enables panic message output for fb cma helper framebuffers.
> 
> Signed-off-by: Noralf Trønnes <noralf at tronnes.org>
> ---
>  drivers/gpu/drm/drm_fb_cma_helper.c | 10 ++++++++++
>  1 file changed, 10 insertions(+)
> 
> diff --git a/drivers/gpu/drm/drm_fb_cma_helper.c
> b/drivers/gpu/drm/drm_fb_cma_helper.c index 1fd6eac..2f1b012 100644
> --- a/drivers/gpu/drm/drm_fb_cma_helper.c
> +++ b/drivers/gpu/drm/drm_fb_cma_helper.c
> @@ -126,9 +126,19 @@ int drm_fb_cma_create_handle(struct drm_framebuffer
> *fb, }
>  EXPORT_SYMBOL(drm_fb_cma_create_handle);
> 
> +static void *drm_fb_cma_panic_vmap(struct drm_framebuffer *fb)
> +{
> +	struct drm_fb_cma *fb_cma = to_fb_cma(fb);
> +	struct drm_gem_cma_object *cma_obj = fb_cma->obj[0];
> +
> +	/* A PRIME imported buffer will not have it's vaddr set. */

Does this mean that, if the framebuffer that happens to be displayed when a 
panic occurs is imported, no message will be printed ? I understand how 
supporting such cases is difficult, but I'm wondering how we could proceed to 
ensure that a panic can be displayed in most (if not all) cases.

Similarly, it looks like your code only handles single-planar formats, but 
there's no explicit check for that in patch 1/3. The easiest fix is to reject 
multi-planar framebuffers, but that would again result in silent panics in 
some cases.

> +	return cma_obj ? cma_obj->vaddr : NULL;

Can cma_obj be NULL here ? I thought that framebuffer objects were always 
created with at least one GEM object.

> +}
> +
>  static struct drm_framebuffer_funcs drm_fb_cma_funcs = {
>  	.destroy	= drm_fb_cma_destroy,
>  	.create_handle	= drm_fb_cma_create_handle,
> +	.panic_vmap	= drm_fb_cma_panic_vmap,
>  };
> 
>  static struct drm_fb_cma *drm_fb_cma_alloc(struct drm_device *dev,

-- 
Regards,

Laurent Pinchart